Load Libraries & Report Versions

library(Seurat)
library(dplyr)
library(ulimit) # devtools::install_github("krlmlr/ulimit")

## Error in library(ulimit): there is no package called 'ulimit'

library(benchmarkme)

## Error in library(benchmarkme): there is no package called 'benchmarkme'

library(gridExtra)
library(knitr) 
library(plotly)
library(ggplot2)
library(viridis)
library(reshape2)
library(shiny) 
library(ggrepel)
library(DT) 
library(ComplexHeatmap); #BiocManager::install("ComplexHeatmap") 
 
# install.packages('devtools')
# devtools::install_github('talgalili/heatmaply')
## Install Bioconductor
#  if (!requireNamespace("BiocManager"))
#     install.packages("BiocManager")
# BiocManager::install(c("biomaRt"))
library(biomaRt)
# BiocManager::install(c("DESeq2"))
library(DESeq2)
# library(snow); #BiocManager::install("Rmpi") #NOTE: different lib name than install name (snow vs Rmpi)
 
createDT <- function(DF, caption="", scrollY=500){
  data <- DT::datatable(DF, caption=caption,
    extensions =  'Buttons',
    options = list( dom = 'Bfrtip', 
                    buttons = c('copy', 'csv', 'excel', 'pdf', 'print'), 
                    scrollY = scrollY, scrollX=T, scrollCollapse = T, paging = F,  
                      columnDefs = list(list(className = 'dt-center', targets = "_all"))
    )
  ) 
   return(data)
}
 
# Useful Seurat functions
## Seurat::FindGeneTerms() # Enrichr API
## Seurat::MultiModal_CCA() # Integrates data from disparate datasets (CIA version too)

sessionInfo()

## R version 3.5.1 (2018-07-02)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: CentOS release 6.9 (Final)
## 
## Matrix products: default
## BLAS/LAPACK: /hpc/packages/minerva-common/intel/parallel_studio_xe_2018/compilers_and_libraries_2018.1.163/linux/mkl/lib/intel64_lin/libmkl_gf_lp64.so
## 
## locale:
## [1] C
## 
## attached base packages:
##  [1] parallel  stats4    grid      stats     graphics  grDevices utils    
##  [8] datasets  methods   base     
## 
## other attached packages:
##  [1] DESeq2_1.20.0               SummarizedExperiment_1.10.1
##  [3] DelayedArray_0.6.1          BiocParallel_1.14.2        
##  [5] matrixStats_0.53.1          Biobase_2.40.0             
##  [7] GenomicRanges_1.32.4        GenomeInfoDb_1.16.0        
##  [9] IRanges_2.14.10             S4Vectors_0.18.3           
## [11] BiocGenerics_0.26.0         biomaRt_2.36.1             
## [13] ComplexHeatmap_1.18.0       DT_0.4                     
## [15] ggrepel_0.8.0               shiny_1.1.0                
## [17] reshape2_1.4.3              viridis_0.5.1              
## [19] viridisLite_0.3.0           plotly_4.7.1               
## [21] knitr_1.20                  gridExtra_2.3              
## [23] dplyr_0.7.6                 Seurat_2.3.4               
## [25] Matrix_1.2-14               cowplot_0.9.3              
## [27] ggplot2_3.0.0              
## 
## loaded via a namespace (and not attached):
##   [1] snow_0.4-3             backports_1.1.2        circlize_0.4.4        
##   [4] Hmisc_4.1-1            plyr_1.8.4             igraph_1.2.1          
##   [7] lazyeval_0.2.1         splines_3.5.1          digest_0.6.17         
##  [10] foreach_1.4.4          htmltools_0.3.6        lars_1.2              
##  [13] gdata_2.18.0           magrittr_1.5           checkmate_1.8.5       
##  [16] memoise_1.1.0          cluster_2.0.7-1        mixtools_1.1.0        
##  [19] ROCR_1.0-7             annotate_1.58.0        R.utils_2.7.0         
##  [22] prettyunits_1.0.2      colorspace_1.3-2       blob_1.1.1            
##  [25] crayon_1.3.4           RCurl_1.95-4.11        jsonlite_1.5          
##  [28] genefilter_1.62.0      bindr_0.1.1            survival_2.42-6       
##  [31] zoo_1.8-3              iterators_1.0.10       ape_5.1               
##  [34] glue_1.3.0             gtable_0.2.0           zlibbioc_1.26.0       
##  [37] XVector_0.20.0         GetoptLong_0.1.7       kernlab_0.9-26        
##  [40] shape_1.4.4            prabclus_2.2-6         DEoptimR_1.0-8        
##  [43] scales_0.5.0           mvtnorm_1.0-8          DBI_1.0.0             
##  [46] Rcpp_1.0.0             metap_0.9              dtw_1.20-1            
##  [49] xtable_1.8-2           progress_1.2.0         htmlTable_1.12        
##  [52] reticulate_1.7         foreign_0.8-70         bit_1.1-14            
##  [55] proxy_0.4-22           mclust_5.4.1           SDMTools_1.1-221      
##  [58] Formula_1.2-3          tsne_0.1-3             htmlwidgets_1.2       
##  [61] httr_1.3.1             gplots_3.0.1           RColorBrewer_1.1-2    
##  [64] fpc_2.1-11             acepack_1.4.1          modeltools_0.2-22     
##  [67] ica_1.0-2              pkgconfig_2.0.2        XML_3.98-1.12         
##  [70] R.methodsS3_1.7.1      flexmix_2.3-14         nnet_7.3-12           
##  [73] locfit_1.5-9.1         tidyselect_0.2.4       rlang_0.2.1           
##  [76] later_0.7.3            AnnotationDbi_1.42.1   munsell_0.5.0         
##  [79] tools_3.5.1            RSQLite_2.1.1          ggridges_0.5.0        
##  [82] evaluate_0.11          stringr_1.3.1          yaml_2.1.19           
##  [85] bit64_0.9-7            fitdistrplus_1.0-9     robustbase_0.93-1.1   
##  [88] caTools_1.17.1         purrr_0.2.5            RANN_2.6              
##  [91] bindrcpp_0.2.2         pbapply_1.3-4          nlme_3.1-137          
##  [94] mime_0.5               R.oo_1.22.0            hdf5r_1.0.0           
##  [97] compiler_3.5.1         rstudioapi_0.7         png_0.1-7             
## [100] geneplotter_1.58.0     tibble_1.4.2           stringi_1.2.4         
## [103] lattice_0.20-35        trimcluster_0.1-2.1    pillar_1.3.0          
## [106] lmtest_0.9-36          GlobalOptions_0.1.0    data.table_1.11.8     
## [109] bitops_1.0-6           irlba_2.3.2            httpuv_1.4.5          
## [112] R6_2.3.0               latticeExtra_0.6-28    promises_1.0.1        
## [115] KernSmooth_2.23-15     codetools_0.2-15       MASS_7.3-50           
## [118] gtools_3.8.1           assertthat_0.2.0       rprojroot_1.3-2       
## [121] rjson_0.2.20           withr_2.1.2            GenomeInfoDbData_1.1.0
## [124] diptest_0.75-7         doSNOW_1.0.16          hms_0.4.2             
## [127] rpart_4.1-13           tidyr_0.8.1            class_7.3-14          
## [130] rmarkdown_1.10         segmented_0.5-3.0      Rtsne_0.13            
## [133] base64enc_0.1-3

print(paste("Seurat ", packageVersion("Seurat")))

## [1] "Seurat  2.3.4"

Raise Memory Limit

Rstudio has a default memory limit of only 1GB. To override this, detect the true memory available and set a new limit.

RAM <- print(benchmarkme::get_ram())

## Error in loadNamespace(name): there is no package called 'benchmarkme'

## Convert GB to Mib
RAM_Mib <- strsplit(RAM, " ")[[1]][1] %>% as.numeric() * 953.67431640625

## Error in strsplit(RAM, " "): object 'RAM' not found

cat(paste("Available RAM:",RAM))

## Error in eval(quote(list(...)), env): object 'RAM' not found

## Set new memory limit 
ulimit::memory_limit(RAM_Mib) 

## Error in loadNamespace(name): there is no package called 'ulimit'

Load Data

## ! IMPORTANT! Must not setwd to local path when launching on cluster
# setwd("~/Desktop/PD_scRNAseq/")
dir.create(file.path(root,"Data"), showWarnings=F) 
load(file.path(root,"Data/seurat_object_add_HTO_ids.Rdata"))
pbmc <- seurat.obj  
rm(seurat.obj)

pbmc

## An object of class seurat in project RAJ_13357 
##  24914 genes across 22113 samples.

Clean Metadata

Add Metadata

metadata <- read.table(file.path(root,"Data/meta.data4.tsv"))
createDT( metadata, caption = "Metadata")  

## Warning in instance$preRenderHook(instance): It seems your data is too
## big for client-side DataTables. You may consider server-side processing:
## https://rstudio.github.io/DT/server.html

# Make AgeGroups
makeAgeGroups <- function(){
  dim(metadata)
  getMaxRound <- function(vals=metadata$Age, unit=10)unit*ceiling((max(vals)/unit))
  getMinRound <- function(vals=metadata$Age, unit=10)unit*floor((min(vals)/unit)) 
   
  ageBreaks = c(seq(getMinRound(), getMaxRound(), by = 10), getMaxRound()+10)
  AgeGroupsUniq <- c()
  for (i in 1:(length(ageBreaks)-1)){ 
    AgeGroupsUniq <- append(AgeGroupsUniq, paste(ageBreaks[i],ageBreaks[i+1], sep="-")) 
  } 
  data.table::setDT(metadata,keep.rownames = T,check.names = F)[, AgeGroups := cut(Age, 
                                  breaks = ageBreaks, 
                                  right = F, 
                                  labels = AgeGroupsUniq,
                                  nclude.lowest=T)]
  metadata <- data.frame(metadata)
  unique(metadata$AgeGroups)
  head(metadata)
  dim(metadata)
  return(metadata)
}
# metadata <- makeAgeGroups()

pbmc <- AddMetaData(object = pbmc, metadata = metadata)  
# Get rid of any NAs (cells that don't match up with the metadata) 
if(subsetCells==F){
  pbmc <- FilterCells(object = pbmc,  subset.names = "nGene", low.thresholds = 0)
} else {pbmc <- FilterCells(object = pbmc,  subset.names = "nGene", low.thresholds = 0,
                    # Subset for testing 
                    cells.use = pbmc@cell.names[0:subsetCells]
                    )
}  

## Warning in as.list.environment(environment(), all = TRUE): NAs introduced
## by coercion

## Error in 0:subsetCells: NA/NaN argument

Filter & Normalize Data

subsetBiotypes <- function(pbmc, subsetGenes){
  if( subsetGenes!=F ){
    cat(paste("Subsetting genes:",subsetGenes))
    # If the gene_biotypes file exists, import csv. Otherwise, get from biomaRt
    if(file_test("-f", file.path(root,"Data/gene_biotypes.csv"))){
      biotypes <- read.csv(file.path(root,"Data/gene_biotypes.csv"))
    }
    else {
      ensembl <- useMart(biomart="ENSEMBL_MART_ENSEMBL", host="grch37.ensembl.org",
                       dataset="hsapiens_gene_ensembl") 
      ensembl <- useDataset(mart = ensembl, dataset = "hsapiens_gene_ensembl")
      listFilters(ensembl)
      listAttributes(ensembl)   
      biotypes <- getBM(attributes=c("hgnc_symbol", "gene_biotype"), filters="hgnc_symbol",
            values=row.names(pbmc@data), mart=ensembl) 
      write.csv(biotypes, file.path(root,"Data/gene_biotypes.csv"), quote=F, row.names=F)
    } 
    # Subset data by creating new Seurat object (annoying but necessary)
    geneSubset <- biotypes[biotypes$gene_biotype==subsetGenes,"hgnc_symbol"] 
    
    cat(paste(dim(pbmc@raw.data[geneSubset, ])[1],"/", dim(pbmc@raw.data)[1], 
                "genes are", subsetGenes))
    # Add back into pbmc 
    subset.matrix <- pbmc@raw.data[geneSubset, ] # Pull the raw expression matrix from the original Seurat object containing only the genes of interest
    pbmc_sub <- CreateSeuratObject(subset.matrix) # Create a new Seurat object with just the genes of interest
    orig.ident <- row.names(pbmc@meta.data) # Pull the identities from the original Seurat object as a data.frame
    pbmc_sub <- AddMetaData(object = pbmc_sub, metadata = pbmc@meta.data) # Add the idents to the meta.data slot
    pbmc_sub <- SetAllIdent(object = pbmc_sub, id = "ident") # Assign identities for the new Seurat object
    pbmc <- pbmc_sub
    rm(list = c("pbmc_sub","geneSubset", "subset.matrix", "orig.ident")) 
  } 
}

subsetBiotypes(pbmc, subsetGenes)

## Subsetting genes: protein_coding14827 / 24914 genes are protein_coding

pbmc <- FilterCells(object = pbmc, subset.names = c("nGene", "percent.mito"), 
    low.thresholds = c(200, -Inf), high.thresholds = c(2500, 0.05))

pbmc <- NormalizeData(object = pbmc, normalization.method = "LogNormalize", 
    scale.factor = 10000)

# Store the top most variable genes in @var.genes
pbmc <- FindVariableGenes(object = pbmc, mean.function = ExpMean, dispersion.function = LogVMR,
    x.low.cutoff = 0.0125, x.high.cutoff = 3, y.cutoff = 0.5)

# IMPORTANT!: Must set do.par=T and num.cors = n for large datasets being processed on computing clusters
# IMPORTANT!: Use only the var.genes identified by 'FindVariableGenes' as the 'gene.use' arg in 'ScaleData'
## This will greatly reduced the computational load.

# par.Cores <- ifelse(nCores <= 12, 48, nCores)
pbmc <- ScaleData(object = pbmc, genes.use = pbmc@var.genes, vars.to.regress = c("nUMI", "percent.mito"), 
                  do.par = F, num.cores = nCores)

## Regressing out: nUMI, percent.mito

## Warning in RegressOutResid(object = object, vars.to.regress =
## vars.to.regress, : For parallel processing, please set do.par to TRUE.

## 
## Time Elapsed:  23.1420209407806 secs

## Scaling data matrix

pbmc

## An object of class seurat in project RAJ_13357 
##  24914 genes across 19144 samples.

Violin Plots

vp <- VlnPlot(object = pbmc, features.plot = c("nGene", "nUMI", "percent.mito"),nCol = 3, do.return = T) %>% + ggplot2::aes(alpha=0.5)
vp

Find Cell Clusters

We first construct a KNN graph based on the euclidean distance in PCA space, and refine the edge weights between any two cells based on the shared overlap in their local neighborhoods (Jaccard similarity). To cluster the cells, we apply modularity optimization techniques such as the Louvain algorithm (default) or SLM [SLM, Blondel et al., Journal of Statistical Mechanics], to iteratively group cells together, with the goal of optimizing the standard modularity function.

On Resolution
NA

• To further increase speed, you can employ an approximate nearest neighbor search via the RANN package by increasing the nn.eps + parameter. Setting this at 0 (the default) represents an exact neighbor search.
• By default, we perform 100 random starts for clustering and select the result with highest modularity. You can lower this through the n.start parameter to reduce clustering time.

# TRY DIFFERENT RESOLUTIONS
pbmc <- StashIdent(object = pbmc, save.name = "pre_clustering") 
# pbmc <- SetAllIdent(object = pbmc, id = "pre_clustering") 

pbmc <- FindClusters(object = pbmc, reduction.type = "pca", dims.use = 1:10, 
                     resolution = resolution, print.output = T, save.SNN = T, 
                     n.start = 10, nn.eps = 0.5) 

## Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
## 
## Number of nodes: 19144
## Number of edges: 769919
## 
## Running Louvain algorithm...
## Random start: 1
## Iteration: 1
## Modularity: 0.7613
## Iteration: 2
## Modularity: 0.7681
## Iteration: 3
## Modularity: 0.7681
## 
## Random start: 2
## Iteration: 1
## Modularity: 0.7540
## Iteration: 2
## Modularity: 0.7699
## Iteration: 3
## Modularity: 0.7699
## 
## Random start: 3
## Iteration: 1
## Modularity: 0.7661
## Iteration: 2
## Modularity: 0.7707
## Iteration: 3
## Modularity: 0.7707
## 
## Random start: 4
## Iteration: 1
## Modularity: 0.7603
## Iteration: 2
## Modularity: 0.7683
## Iteration: 3
## Modularity: 0.7683
## 
## Random start: 5
## Iteration: 1
## Modularity: 0.7594
## Iteration: 2
## Modularity: 0.7664
## Iteration: 3
## Modularity: 0.7664
## 
## Random start: 6
## Iteration: 1
## Modularity: 0.7620
## Iteration: 2
## Modularity: 0.7695
## Iteration: 3
## Modularity: 0.7701
## Iteration: 4
## Modularity: 0.7701
## 
## Random start: 7
## Iteration: 1
## Modularity: 0.7602
## Iteration: 2
## Modularity: 0.7684
## Iteration: 3
## Modularity: 0.7684
## 
## Random start: 8
## Iteration: 1
## Modularity: 0.7598
## Iteration: 2
## Modularity: 0.7652
## Iteration: 3
## Modularity: 0.7652
## 
## Random start: 9
## Iteration: 1
## Modularity: 0.7621
## Iteration: 2
## Modularity: 0.7679
## Iteration: 3
## Modularity: 0.7679
## 
## Random start: 10
## Iteration: 1
## Modularity: 0.7530
## Iteration: 2
## Modularity: 0.7702
## Iteration: 3
## Modularity: 0.7702
## 
## Maximum modularity in 10 random starts: 0.7707
## Number of communities: 6
## Elapsed time: 9 seconds

PrintFindClustersParams(object = pbmc) 

## Parameters used in latest FindClusters calculation run on: 2019-01-16 13:42:52
## =============================================================================
## Resolution: 0.4
## -----------------------------------------------------------------------------
## Modularity Function    Algorithm         n.start         n.iter
##      1                   1                 10              10
## -----------------------------------------------------------------------------
## Reduction used          k.param          prune.SNN
##      pca                 30                0.0667
## -----------------------------------------------------------------------------
## Dims used in calculation
## =============================================================================
## 1 2 3 4 5 6 7 8 9 10

pbmc <- StashIdent(object = pbmc, save.name = "post_clustering") 

UMAP

Additional UMAP arguments detailed here: https://umap-learn.readthedocs.io/en/latest/api.html#module-umap.umap_

pbmc <- RunUMAP(object = pbmc, reduction.use = "pca", dims.use = 1:10, verbose=TRUE, num_threads=nCores) # , num_threads=0 
# Plot results
DimPlot(object = pbmc, reduction.use = 'umap')

t-SNE

As input to the tSNE, we suggest using the same PCs as input to the clustering analysis, although computing the tSNE based on scaled gene expression is also supported using the genes.use argument.

** Important!**: Specify num_threads=0 in ‘RunTSNE’ to use all available cores.

“FItSNE”, a new fast implementation of t-SNE, is also available through RunTSNE. However FItSNE must first be setup on your computer.

labSize <- 6 

pbmc <- RunTSNE(object=pbmc,  reduction.use = "pca", dims.use = 1:10, do.fast = TRUE,
                tsne.method = "Rtsne", num_threads=nCores, verbose=T) #   FItSNE

## Read the 19144 x 10 data matrix successfully!
## Using no_dims = 2, perplexity = 30.000000, and theta = 0.500000
## Computing input similarities...
## Normalizing input...
## Building tree...
##  - point 0 of 19144
##  - point 10000 of 19144
## Done in 9.44 seconds (sparsity = 0.006869)!
## Learning embedding...
## Iteration 50: error is 104.835938 (50 iterations in 19.62 seconds)
## Iteration 100: error is 104.709901 (50 iterations in 25.21 seconds)
## Iteration 150: error is 101.632348 (50 iterations in 21.25 seconds)
## Iteration 200: error is 101.580125 (50 iterations in 22.61 seconds)
## Iteration 250: error is 101.472564 (50 iterations in 20.42 seconds)
## Iteration 300: error is 4.413932 (50 iterations in 13.54 seconds)
## Iteration 350: error is 4.166479 (50 iterations in 14.89 seconds)
## Iteration 400: error is 4.016408 (50 iterations in 15.17 seconds)
## Iteration 450: error is 3.905556 (50 iterations in 15.98 seconds)
## Iteration 500: error is 3.818781 (50 iterations in 16.77 seconds)
## Iteration 550: error is 3.750168 (50 iterations in 16.79 seconds)
## Iteration 600: error is 3.693278 (50 iterations in 16.65 seconds)
## Iteration 650: error is 3.645445 (50 iterations in 16.63 seconds)
## Iteration 700: error is 3.604156 (50 iterations in 16.93 seconds)
## Iteration 750: error is 3.568189 (50 iterations in 16.37 seconds)
## Iteration 800: error is 3.536739 (50 iterations in 17.20 seconds)
## Iteration 850: error is 3.509031 (50 iterations in 17.36 seconds)
## Iteration 900: error is 3.484408 (50 iterations in 17.27 seconds)
## Iteration 950: error is 3.462351 (50 iterations in 17.38 seconds)
## Iteration 1000: error is 3.443001 (50 iterations in 17.65 seconds)
## Fitting performed in 355.69 seconds.

# note that you can set do.label=T to help label individual clusters
TSNEPlot(object = pbmc, do.label=T, label.size = labSize, do.return=T) 

# ifelse(interactive, tp %>% ggplotly() %>% toWebGL() %>% print(), print(tp))

Top Biomarker Plot

fp <- FeaturePlot(object = pbmc, features.plot = top1$gene, cols.use = c("grey", "purple"), 
    reduction.use = "tsne", nCol = nCols, do.return = T)

top5 <- pbmc.markers %>% group_by(cluster) %>% top_n(5, avg_logFC)
# setting slim.col.label to TRUE will print just the cluster IDS instead of
# every cell name
DoHeatmap(object = pbmc, genes.use = top5$gene, slim.col.label=T, remove.key=T) 

# ifelse(interactive, p %>% ggplotly() %>% toWebGL() %>% print(), print(p))

RidgePlot(pbmc, features.plot = top1$gene,  nCol = nCols, do.sort = F)

## Picking joint bandwidth of 0.207

## Picking joint bandwidth of 0.33

## Picking joint bandwidth of 0.0337

## Picking joint bandwidth of 0.0245

## Picking joint bandwidth of 0.0541

## Picking joint bandwidth of 0.139

top2 <- pbmc.markers %>% group_by(cluster) %>% top_n(2, avg_logFC)

sdp <- SplitDotPlotGG(pbmc, genes.plot = top2$gene, cols.use = c("blue","red"), 
                      x.lab.rot = T, plot.legend = T, dot.scale = 8, do.return = T, grouping.var = "dx")

Markers Dataframe

markerList <- c("CD14", "FCGR3A") 
 
get_markerDF <- function(pbmc, markerList, meta_vars =c("barcode", "dx", "mut","post_clustering", "percent.mito","nGene", "nUMI")){
  exp <- pbmc@scale.data %>% data.frame() 
  marker.matrix <- exp[row.names(exp) %in% markerList, ] 
  marker.matrix$Gene <- row.names(marker.matrix)
  markerMelt <- reshape2:::melt.data.frame(marker.matrix, id.vars = "Gene", variable.name = "Cell",value.name = "Expression") 
  metaSelect <-  pbmc@meta.data[,meta_vars] 
  markerDF <- merge(markerMelt,metaSelect, by.x="Cell", by.y="barcode") 
  return(markerDF)
}
markerDF <- get_markerDF(pbmc, markerList)
createDT(markerDF, caption = "Known Marker Expression")

## Warning in instance$preRenderHook(instance): It seems your data is too
## big for client-side DataTables. You may consider server-side processing:
## https://rstudio.github.io/DT/server.html

Marker ANOVAs + Boxplots

# Explore expression differences between groups
marker_vs_metadata <- function(markerDF, meta_var){ 
  # Create title from ANOVA summary
  ANOVAtitle <- function(markerDF, marker){
      nTests <- length(unique(markerDF$Gene))
      res <- anova(lm(data = subset(markerDF, Gene==marker), 
                      formula = Expression ~ eval(parse(text=meta_var))))
      
      title <-paste(paste("ANOVA (",marker, " vs. ",meta_var, ")", sep=""), 
                    ": p=",round(res$`Pr(>F)`,3), 
                    ", F=",round(res$`F value`,3), 
        ifelse(res$`Pr(>F)`<.05/nTests,"(Significant**)",
               "(Non-significant)") ) 
  }
  
  title = ""
  for (marker in unique(markerDF$Gene) ){
    cat(marker)
    title <- paste(title, "\n", ANOVAtitle(markerDF, marker))
  } 
  
  ggplot(markerDF, aes(x=eval(parse(text=meta_var)), y=Expression, fill= Gene)) + 
    geom_boxplot() +  
    labs(title = title, x=meta_var) +
    theme(plot.title = element_text( size=10)) +
    scale_fill_manual(values=c("brown", "slategray"))
}

marker_vs_metadata(markerDF, "dx")

## FCGR3A

marker_vs_metadata(markerDF, "mut") 

## FCGR3A

Defining Cell-types

identify_cellTypes_by_biomarkers <- function(pbmc.markers, topN_search=5){
  top <- pbmc.markers %>% group_by(cluster) %>% top_n(topN_search, avg_logFC) 
  clust_cellTypes <- list()
  for (clust in top$cluster){ 
    clustSub <- top[top5$cluster==clust, ]
    CD16_logFC <- subset(clustSub, gene=="CFD")$avg_logFC 
    
    cellType <- ifelse(sum(markerList %in% clustSub$gene), # Both CD14 and CD16? Great, keep going
           ifelse(CD16_logFC == abs(CD16_logFC), "CD14++/CD16+",  # But does CD16 have pos logfC? If so, then it's "CD14++/CD16+"
                  "CD14++/CD16--"), # Otherwise, it means it means CD16 logFC is neg, meaning "CD14++/CD16--"
           NA) # If it's none of these, it's an undefined cell type 
     clust_cellTypes[clust] <- as.factor(cellType)
  }
  newMeta <- pbmc@meta.data
  newMeta["CellType_DGE"] <- plyr::mapvalues(newMeta$post_clustering, names(clust_cellTypes), as.character(clust_cellTypes) )  
  pbmc <- AddMetaData(pbmc, metadata = newMeta)
  return(pbmc)
} 
pbmc <- identify_cellTypes_by_biomarkers(pbmc.markers, 5)

# (Doesn't make sense to do bar plot because whole clusters are defined by their biomarkers)

tSNE_metadata_plot("CellType_DGE")

## t-SNE Metadata plot for  CellType_DGE

Identify Cell Types By Average Expression

• Classification Approach 2
  • Alternatively, you could classify cells by their absolute expression values. E.g. If any cell expresses CD16 more than the average expression of CD16 across all cells, then classify it as CD16+.
  • Drawback: The threshold for classifying a cell as either CD16+ or CD16– is defined rather arbitrarily (the average expression of CD16 in this sample isn’t necessarily biologically meaningful).

# A simplistic way of categorizing cells into CD14++/CD16+ and CD14++/CD16--, 
## is by splitting cells into groups based on whether their expression is 
## higher or lower than the average CD16 expression of all cells.
identify_cellTypes_by_avgExpression <- function(pbmc, markerDF){
  avgMarkerExp <-markerDF %>% group_by(Gene) %>% dplyr::summarise(meanExp = mean(Expression))
  avgMarkerExp <- setNames(avgMarkerExp$meanExp, avgMarkerExp$Gene)
  CD16 <- markerDF[markerDF$Gene=="FCGR3A",]
  CD16_group <- ifelse(CD16$Expression >= avgMarkerExp["FCGR3A"], "CD14++/CD16+", "CD14++/CD16--")
  CD16["CellType_AvgExp"]  <- as.factor(CD16_group)
  # Make sure row order is same before putting back into meta.data
  metaD <- pbmc@meta.data
  newMeta <- merge(metaD, CD16[,c("Cell","CellType_AvgExp")], by.x="barcode", by.y="Cell")
  row.names(newMeta) <- row.names(metaD)
  pbmc <- AddMetaData(pbmc, metadata = newMeta)
  return(pbmc)
}
pbmc <- identify_cellTypes_by_avgExpression(pbmc, markerDF)
 
# Get proportions of cell types in each cluster
cluster_proportions <- pbmc@meta.data %>% group_by(CellType_AvgExp, post_clustering) %>% 
  tally() %>%
  group_by(post_clustering, CellType_AvgExp) %>%
  mutate(percentTotal = n/sum(n)*100)

ggplot(cluster_proportions, aes(x=post_clustering, y=percentTotal, fill=CellType_AvgExp)) + geom_col(position = "fill") +  
  labs(title="Proportions of Cell-types per Cluster: \n CellType_AvgExp", 
       x="Cluster", y="Cell Type / Total Cells") +
  scale_fill_manual(values=c("brown", "slategray"))

tSNE_metadata_plot("CellType_AvgExp")

## t-SNE Metadata plot for  CellType_AvgExp

Known Biomarkers: Boxplot

ggplot(data = markerDF, aes(x=Cluster, y=Expression, fill=Gene)) %>% 
  + geom_boxplot(alpha=0.5) %>% + scale_fill_manual(values=c("purple", "turquoise"))  

Known Biomarkers: tSNE

#, results = 'hide', fig.show='hide'
expressionTSNE <- function(pbmc, marker, colors=c("grey", "red")){
   FeaturePlot(object = pbmc, features.plot = marker, cols.use = colors, 
    reduction.use = "tsne", nCol=2, do.return = T, dark.theme = T)[[1]]
  # p <- ifelse(interactive, p %>% ggplotly() %>% toWebGL(), print(p)) 
}
 plot_grid(expressionTSNE(pbmc, markerList[1]),
 expressionTSNE(pbmc, markerList[2], colors=c("grey", "green")))

Label Clusters by DGE Biomarkers

current.cluster.ids <- unique(pbmc.markers$cluster) #c(0, 1, 2, 3, 4, 5, 6, 7)
top1 <- pbmc.markers %>% group_by(cluster) %>% top_n(1, avg_logFC)
new.cluster.ids <- top1$gene #c("CD4 T cells", "CD14+ Monocytes", "B cells", "CD8 T cells", "FCGR3A+ Monocytes", "NK cells", "Dendritic cells", "Megakaryocytes")

pbmc@ident <- plyr::mapvalues(x = pbmc@ident, from = current.cluster.ids, to = new.cluster.ids)
TSNEPlot(object=pbmc, do.label=T, pt.size=0.5, do.return=T) 

# ifelse(interactive, p %>% ggplotly() %>% toWebGL() %>% print(), print(p))